Using determiners as contextual cues in sentence comprehension:
A comparison between younger and older adults
Nazbanou Nozari ([email protected])
Department of Neurology, Johns Hopkins University, 1629 Thames Street
Baltimore, MD 21231, USA
Daniel Mirman ([email protected])
Department of Psychology, Drexel University, 3141 Chestnut St
Philadelphia, PA 19104, USA
Abstract
Younger adults use both semantic and phonological cues to
quickly and efficiently localize the referent during sentence
comprehension. While some behavioral studies suggest that
older adults use contextual information even more strongly
than younger adults, ERP studies have shown that this
population, as a group, is less apt at using contextual semantic
cues to predict upcoming words. The current study extends
the investigation of contextual cue processing in auditory
sentence comprehension beyond semantic cue processing, by
comparing younger and older adults in their ability to use
phonological cues in indefinite articles (a/an) to localize the
referent in an eye-tracking visual world paradigm. Our results
suggest that both age groups use such phonological
information for referent localization, but with different
timelines: younger adults use the cues to anticipate an
upcoming word, whereas older adults show delayed cue
processing after the target word has been spoken. Together
with findings from semantic context processing, these results
support a model of sentence comprehension in which the use
of contextual cues continues with aging, but is no longer as
efficient as in the young system for anticipatory word
retrieval.
Keywords: aging; sentence comprehension;
determiners; indefinite articles; eye-tracking.
context;
Introduction
Do older adults process sentences differently from younger
adults? The answer to this question is not only interesting
from the viewpoint of understanding the language system,
but also from the more general perspective of how aging
changes the principles of cognitive processing. One such
principle is the use of contextual information to facilitate
processing of an upcoming stimulus. A large body of
research suggests that individuals routinely use contextual
information in both reading and listening; for example, upon
hearing “The boy will eat the …” listeners quickly look at
the edible object among all other objects on the screen (e.g.,
Altmann & Kamide, 1999). Similarly, upon reading “The
day was breezy so the boy went outside to fly ...”
participants anticipate the word “kite” and its appropriate
determiner “a”, showing not only anticipation of the lexical
representation but also of its phonological form (Delong et
al., 2005, 2012). Do older adults also use contextual
information?
Some behavioral studies suggest that older adults use
sentential context to process a target word as much as, or
even more than, younger adults, especially when the target
is presented amid visual or auditory noise (e.g., Madden,
1988; Pichora-Fuller, Schneider, & Daneman, 1995).
However, more recent ERP studies have shown that older
adults are less likely to use contextual information during
sentence comprehension to anticipate the upcoming word
(e.g., Federmeier, Kutas, & Schul, 2010; Federmeier,
McLennan, Ochoa, & Kutas, 2002; Wlotko, Federmeier, &
Kutas, 2012).
Most of these ERP studies have focused on N400, a
negative potential that reflects implicit aspects of semantic
access, and is larger in magnitude for words that contradict
contextually-induced expectations (Kutas & Federmeier,
2000). For example, Federmeier, Van Petten, Schwartz, and
Kutas (2003) found that N400 in response to conflict
between a word and the sentential context was delayed by
over 200 ms in older, compared to younger, adults.
Subsequent work endorsed this finding, by showing that
N400 reduction in response to strongly-constraining context
was smaller and significantly delayed in older adults,
especially those with lower reading spans (Federmeier &
Kutas, 2005). Federmeier et al. (2010) further showed that
the differences in using contextual cues was still visible
between younger and older adults, even when the working
memory load of the task was minimal, although older adults
with higher verbal fluency scores showed ERP patterns that
more closely resembled that of younger adults.
The ERP studies reviewed above suggest that older
adults, as a group, are less likely to use contextual
information to pre-activate potential target representations.
To determine whether this age group showed any trace of
context use for pre-activating linguistic information,
DeLong, Groppe, Urbach, and Kutas, (2012) measured
N400 responses to indefinite articles “a” and “an” for
upcoming nouns whose cloze probability given the context
was parametrically manipulated to be between 0 and 100%
(e.g., “Dale was very sorry and knew he owed Mary a
check/an apology for what he had done.”). The main finding
was that N400 in response to nouns showed graded
sensitivity to nouns’ cloze probability in both younger and
older adults, but these N400 correlations had a later onset
1
1193
and lasted longer in the older adults (see also Wlotko et al.,
2012). Importantly, unlike younger adults, older adults did
not show any effects on the articles, indicating that at that
early point they had not yet anticipated the phonological
form of the upcoming word. A second finding of the study
was a prolonged increased frontal positivity to less likely
nouns in young adults and a subset of older adults with high
verbal fluency. Together, these findings led the authors to
propose that older adults may in fact use contextual cues for
pre-processing of the linguistic representations, even though
the delayed timeline of this effect might lead one to
conclude otherwise.
We test this possibility by comparing younger and older
adults in a visual world eye-tracking paradigm. Participants
viewed a scene of four objects, while listening to sentences
such as “She will see a/an cherry/apple”. On the
Experimental trials, the indefinite articles “a” or “an”
unambiguously cued the target, because the other three
objects all had names that started with a phoneme that was
incompatible with that article (e.g., for “an” the target would
be “apple”, and the three distractors would be “baby”,
“piano”, and “duck”). On Control trials, participants heard
similar sentences but with the definite article “the”, which
provides weaker cues to the target identity. If listeners use
indefinite articles as cues, they should be able to fixate the
target faster on the Experimental, compared to the Control,
trials.
Importantly, “a” and “an” are semantically identical, and
the disambiguating information they carry is phonological.
Past research has suggested that younger adults use
phonological cues to anticipate potential targets (e.g.,
Allopenna, Magnuson, & Tanenhaus, 1998; Zwitserlood &
Schriefers, 1995). For example, upon hearing the onset /b/
listeners consider all the words that start with /b/ as possible
referents (the cohort effect; Allopenna et al., 1998). Are
phonological cues in the articles also used as cues?
Interestingly, articles may not be fully processed in sentence
comprehension. Readers often skip over articles in reading
(O’Regan, 1979) and children older than age 4 show no
disruption in sentence comprehension when an
inappropriate article is used (Zangl & Fernald, 2007;
McNamara, Carter, McIntosh, & Gerken, 1998). Thus
articles, while potentially valuable phonological cues, may
be skipped without causing much harm to comprehension.
Thus, if older adults are less likely to process contextual
cues, articles would be an excellent test bed.
The current experiment investigated whether indefinite
articles are used as phonological cues to locate the referent
during auditory sentence comprehension in younger and
older adults, and whether the two groups process such cues
differently.
Methods
Participants
Twelve younger adults (six females, mean age = 19.5, SE =
0.4 years), and twelve older adults (five females, mean age
= 63.7, SE = 2.3 years) participated in the study in exchange
for course credit or payment. All participants were native
speakers of English. Older adults were tested on the MiniMental State Exam (Folstein, Folstein, & McHugh, 1975)
and all scored within normal range.
Materials
Visual stimuli were presented as 300×300 pixel pictures of
black and white line-drawings taken from either the IPNP
corpus (Szekely et al., 2004), or from Snodgrass &
Vanderwart (1980). Targets were 60 common nouns, half
beginning with a vowel, and half with a consonant. There
were no significant differences between the items in the two
groups in frequency (t(57) = 1.22 , p = .23; reported as
frequency per million words, from SUBTLEX (Brysbaert &
New, 2009)), number of syllables (t(58) = -1.37, p = .18),
and number of phonemes (t(58) = .27, p = .79). Each item
appeared once as the target in the Experimental condition
(“an apple”), once as the target in the Control condition
(“the apple”), and six more times as distractor in trials with
other target nouns. One hundred and twenty sentences with
the structure “She will see [article][target].” were recorded
by a native English speaker at 44.1 kHz. All sentences were
recorded naturally, without word splicing. This was
necessary because the pronunciation of “the” could change
depending on whether the following noun starts with a
vowel or a consonant. Therefore, “a” and “an” should each
have their own proper baseline of “the”; splicing would
have removed this natural variability in “the” pronunciation
and provided a biased baseline. In the recorded materials,
there was no significant difference between the duration of
the determiners and their paired “the” controls (t(29) = .58,
p = .56 for “a” vs. “the”; t(29) = 1.15, p = .23 for “an” vs.
“the”).
Apparatus
Participants were seated approximately 25 inches away from
a 17-inch monitor with the resolution set to 1024×768 dpi.
Stimuli were presented using E-Prime Professional, Version
2.0 software (Psychology Software Tools, Inc.,
www.pstnet.com). A remote Eyelink 1000 eye-tracker
recorded participants’ monocular gaze position at 250 Hz.
Procedure
Participants were instructed to “listen and look at the
pictures” (no response was required). Each trial began with
a 1375 ms preview. In the first 1000 ms, the four line2
1194
drawings were presented in the four corners, and in the last
375 ms a shrinking red dot appeared at the center to draw
the gaze back to the central location. After the preview, the
sentence was presented through speakers at a comfortable
listening volume. The position of the four pictures was
randomized on every trial. Participants first completed six
practice trials, and then moved on to the experiment. Two
blocks, each containing 60 intermixed “a”, ”an” and “the”
trials were administered, with a break in between. No
picture was repeated as the target within the same block.
The first analysis addressed four questions: (a) Do
phonological cues facilitate target localization? (b) Do both
younger and older adults both use these cues? (c) Are there
differences in using “a” and “an” cues? And (d) Do the two
age groups both show these differences? Table 1 shows the
full model results.
Table 1- Results of the GCA first analysis.
Results
Figure 1 shows the proportion of fixation (±SE) on the
target for the Experimental (a/an) and Control (the)
conditions, separated by article type and age group.
Statistical analyses were performed using the Growth Curve
Analysis method (GCA; Mirman, 2014), a variant of
multilevel regression (or hierarchical linear modeling) that
uses orthogonal polynomials to capture the curvilinear
pattern of fixation proportions over time. Effects of the
variables of interest on the polynomial terms provide a way
to quantify and evaluate those effects on statistically
independent (i.e., orthogonal) aspects of the fixation
proportions trajectory. Data were contrast coded and
centered, and unless otherwise specified, the overall target
fixation trajectory was modeled with a cubic polynomial.
Dependent variables included Age (young vs. old),
Condition (definite article “the” vs. indefinite articles
“a”/”an”) and ArtType (Article Type; type of the indefinite
article: “a” vs. “an”). The interaction between these
variables and polynomial terms were explored over the
intercept, linear and quadratic terms, as higher order
interactions are difficult to interpret (see Mirman et al.,
2014 for a full discussion). The cubic term was, however,
entered in both the fixed the random effect structures to
obtain the best fit to the data.
Figure 1: Proportion of fixations (±SE) to the target in
Experimental (a/an) vs. control (the) conditions, separately
for each article and each age group.
The effect of Condition was significant on the intercept,
linear and quadratic terms, showing that the phonological
cues provided by the indefinite articles reliably facilitated
localization of the target compared to the less
phonologically-informative “the”. This effect was not
reliably different between older and younger adults, as
evidenced by non-significant Condition*Age interactions on
all three polynomial terms, implying that both age groups
used these cues. We then asked if facilitation was more
pronounced for one type of the definite article over the
other. The interaction between Condition and ArtType was
significant over two of the three polynomial terms (intercept
and quadratic), indicating that the article “an” was more
facilitatory than the article “a”, when each article was
compared to its own baseline “the”. Finally, we asked
3
1195
whether younger and older adults differed in their
processing of “a” vs. “an” articles. We found a marginal
effect of the three-way interaction (Condition by ArtType
by Age) on the intercept, hinting at a possible difference.
In summary, the results of this analysis provided strong
evidence that both younger and older adults used
phonological cues associated with the definite articles to
more efficiently locate the visual target, but also suggested
that there might be differences in how the two age groups
used such cues. These differences were explored in the next
set of analyses.
The second set of analyses focused on separate
exploration of facilitation in “an” and “a”. The model
structure was similar to that of the first analysis, except that
ArtType was removed since only subsets of data containing
either “an” or “a” were analyzed by each model. Table 2
presents the results of the analysis on the “a” dataset.
Neither the effect of Condition, nor its interaction with age,
was significant on any of the polynomial terms.
quadratic term (coefficient = -0.018; SE = 0.006; t = 3.260; p = 0.003). This interaction suggests that the
anticipatory effect of phonological cue was reliably more
pronounced in the younger, compared to the older, adults
(See Fig. 2).
Late time window. This time window was chosen to reflect
the opposite of an anticipatory process, namely to explore if
phonological cues were still being used even when the target
noun had been completely spoken and there was no longer
any ambiguity about the referent. To this end, we chose a
time window starting at the end of the noun (920 ms after
the article onset) until the point where fixations plateaued
(1400 ms after the article onset). This analysis revealed a
marginal main effect of Condition on the intercept
(coefficient = -0.033, SE = 0.016; t = -2.058; p = 0.052), as
well as a significant interaction between Condition and Age
on the quadratic term (coefficient = 0.028; SE = 0.011; t =
2.478; p = 0.018) and a marginal effect of this interaction on
the intercept (coefficient = -0.028; SE = 0.016; t = -1.782; p
= 0.089). Importantly, the direction of this interaction was
the opposite of that found in the early time window analysis,
showing that in this late time window, older adults
continued to rely on the earlier phonological cues from the
article to find the referent, considerably more than younger
adults (see Fig. 2).
Table 2- Results of the GCA on the “a” subset.
Figure 2: Proportion of fixations (±SE) on the target in
Experimental (an) vs. control (the) conditions in each age
group. Gray rectangles mark the two analysis time windows.
Table 3 shows the results of the analysis on the “an”
subset. The effect of Condition was significant on the
intercept and quadratic terms, showing that participants used
“an” as a more reliable cue to locate the target compared to
“the”. Importantly, the interaction between Condition and
age was also significant on the linear term. These results
indicate that while, as a group, younger and older adults
both use “an” as an informative cue for finding the referent,
they do so differently. The next set of analyses further
explored these differences by focusing on early and late
time windows.
Early time window. This analysis focused on a narrow
window of 300 ms, starting 200 ms after the presentation of
the article (to allow for planning and execution of an eye
movement), and ending 240 ms after the onset of the word.
This window best captures early use of phonological cues
for anticipatory target localization. We found a main effect
of Condition on the intercept, such that “an” facilitated
target fixation compared to “the” (coefficient = -0.022, SE =
0.009; t = -2.471; p = 0.022). Critically, we also found a
reliable interaction between Condition and Age on the
In summary, the second set of analyses showed that (1)
the effect of phonological cue was stronger in “an” than in
“a”, (2) both younger and older adults used “an” as a cue to
localize the visual referent, and (3) the two age groups used
4
1196
the cue with notably different timelines; early in younger
and late in older adults.
Table 3- Results of the GCA on the “an” subset.
General Discussion
Our results suggest that articles were indeed used as
informative contextual cues for localizing the referent
during online sentence comprehension regardless of age.
This finding suggests that while comprehension does not
critically depend on articles, listeners do use these function
words as cues. However, this benefit was limited to the less
frequent article “an”. Note that each article was compared to
its corresponding control condition (e.g., “a cherry” is
compared to “the cherry” and “an apple” to “the apple”),
thus subtle differences in the pronunciation of “the” when
followed by a consonant or a vowel were controlled for. The
most likely reason for “an” being used as a more prominent
cue than “a”, is that the subset of words that “an” cues (i.e.,
words that start with a vowel) is much smaller than the
subset of words that are cued by “a”. Also in modern
American English, the use of “a” along with a noun that
starts with a vowel, though infelicitous, is not uncommon,
especially in disfluent speech (e.g., “He’s a um… artist!”),
decreasing the validity of “a” as a unique cue.
Critical for our investigation was that older adults too
showed evidence of employing these article cues for
locating the referent. This finding complements the
literature on contextual cue processing in older adults,
which has almost entirely focused on the processing of
semantic context. Note that studies that investigate ERP
responses on the article (e.g., Delong et al., 2005, 2012),
investigate retrieval of the phonological form in response to
semantic cues in the sentence, e.g., retrieving the word
“kite” or “airplane” (and subsequently the appropriate
article a/an), given the context “The day was breezy so the
boy went outside to fly ...”. Thus these studies follow a
different goal from the present study, which tested whether
phonological cues associated with the articles themselves
are used by listeners. Our results suggest that older adults do
use such phonological cues, even on function words that
past research suggests are not critical for sentence
comprehension (e.g., McNamara et al., 1998; Zangl &
Fernald, 2007).
However, our analyses of early and late time windows
revealed intriguing differences in the timeline of cue
processing between younger and older adults. While
younger participants used phonological article cues in an
anticipatory fashion, i.e., to locate the target before the
target name was spoken, older adults showed no evidence of
anticipatory cue use. On the other hand, older adults showed
continued advantage for the target when it followed “an”,
compared to “the”, in the late time window, after the noun
had been spoken. In this late time window, younger adults
showed no difference in fixating targets that were preceded
by either article type, presumably because the noun
information was enough for unequivocally localizing the
referent in both conditions.
The delay in the processing of phonological article cues in
older adults mirrors the reports of delayed N400 effect
during semantic context processing (e.g., Delong et al.,
2012; Federmeier et al., 2003; 2010; Wlotko & Federmeier,
2012; Wlotko et al., 2012). Moreover, older adults’
continued processing of such cues in a late time window is
compatible with Delong et al.’s (2012) report of prolonged
N400 in this population (see also Wlotko et al., 2012). Our
results add to these findings in three ways. First, the delay in
processing contextual cues in older adults is not limited to
semantic cues. This shows that the issue does not stem from
specific dynamics within the lexical-semantic system, and is
instead be better explained by models of general slowing in
cognitive processing with aging (e.g., Myerson et al., 1990).
Importantly, this slowing does not encompass automatic
aspects of processing, such as activation of lexical
representations through related representations (e.g.,
Federmeier et al., 2003; see Burk & Shafto, 2008, for a
review), but seems to be specific to operations that require
active processing and control.
The second way that the current results add to the findings
of ERP studies is that it corroborates those findings using a
measure that is sensitive enough to capture subtle
differences in the timeline of cognitive processing, while
still capturing overt behavior. We intentionally used articles
which, according to past research, were dispensable to
comprehension to show that even such subtle cues can be
used in older adults and have measurable behavioral
consequences. Finally, the sentences used in the current
experiment entailed no anomaly or syntactic complexity,
thus providing a straight-forward test of contextual cue use
in sentence comprehension.
5
1197
In summary, these results, together with those of past
studies, support a model of sentence comprehension in
which the use of semantic and phonological contextual cues
to process upcoming words continues in old age, but the
efficiency of using such cues for anticipatory word retrieval
decreases considerably with aging.
Acknowledgments
The authors to thank Allison Britt and Kristen Graziano for
their help with recordings and data collection. This research
was funded in part by NIH grant R01DC010805 to DM and
by the Moss Rehabilitation Research Institute.
References
Allopenna, P. D., Magnuson, J. S., & Tanenhaus, M. K.
(1998). Tracking the time course of spoken word
recognition using eye movements: Evidence for
continuous mapping models. Journal of Memory and
Language, 38(4), 419–439.
Brysbaert, M., & New, B. (2009). Moving beyond Ku\vcera
and Francis: A critical evaluation of current word
frequency norms and the introduction of a new and
improved word frequency measure for American English.
Behavior Research Methods, 41(4), 977–990.
Burke, D. M., & Shafto, M. (2008). Language and aging. In
F. I. M. Craik & T. A. Salthouse (Eds.). Handbook of
aging and cognition (Vol. 3, pp. 373–444). New York:
Psychology Press.
DeLong, K. A., Groppe, D. M., Urbach, T. P., & Kutas, M.
(2012). Thinking ahead or not? Natural aging and
anticipation during reading. Brain and Language, 121(3),
226–239.
DeLong, K. A., Urbach, T. P., & Kutas, M. (2005).
Probabilistic word pre-activation during language
comprehension inferred from electrical brain activity.
Nature Neuroscience, 8(8), 1117–1121.
Federmeier, K. D., & Kutas, M. (2005). Aging in context:
Age-related changes in context use during language
comprehension. Psychophysiology, 42(2), 133–141.
Federmeier, K. D., Kutas, M., & Schul, R. (2010). Agerelated and individual differences in the use of prediction
during language comprehension. Brain and Language,
115(3), 149–161.
Federmeier, K. D., McLennan, D. B., Ochoa, E., & Kutas,
M. (2002). The impact of semantic memory organization
and sentence context information on spoken language
processing by younger and older adults: An ERP study.
Psychophysiology, 39(2), 133–146.
Federmeier, K. D., Van Petten, C., Schwartz, T. J., & Kutas,
M. (2003). Sounds, words, sentences: age-related changes
across levels of language processing. Psychology and
Aging, 18(4), 858.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975).
“Mini-mental state”: a practical method for grading the
cognitive state of patients for the clinician. Journal of
Psychiatric Research, 12(3), 189–198.
Kutas, M., & Federmeier, K. D. (2000). Electrophysiology
reveals semantic memory use in language comprehension.
Trends in Cognitive Sciences, 4(12), 463–470.
Madden, D. J. (1988). Adult age-differences in the effects of
sentence context and stimulus degradation during visual
word recognition. Psychology and Aging, 3(2), 167–172.
McNamara, M., Carter, A., McIntosh, B., & Gerken, L.
(1998). Sensitivity to grammatical morphemes in children
with specific language impairment.Journal of Speech,
Language, and Hearing Research, 41(5), 1147-1157.
Mirman, D. (2014). Growth Curve Analysis and
Visualization Using R. Florida, USA: Chapman &
Hall/CRC Press.
Myerson, J., Hale, S., Poon, L. W., Wagstaff, D., & Smith,
G. A. (1990). The information-loss model – A
mathematical-theory of age-related cognitive slowing.
Psychological Review, 97(4), 475–487.
O’Regan, K. (1979). Saccade size control in reading:
Evidence for the linguistic control hypothesis. Perception
& Psychophysics, 25(6), 501–509.
Pichora-Fuller, M. K., Schneider, B. A., & Daneman, M.
(1995). How young and old adults listen to and remember
speech in noise. The Journal of the Acoustical Society of
America, 97(1), 593–608.
Snodgrass, J. G., & Vanderwart, M. (1980). A standardized
set of 260 pictures: norms for name agreement, image
agreement, familiarity, and visual complexity. Journal of
Experimental Psychology: Human Learning and Memory,
6(2), 174.
Stine-Morrow, E. A., Miller, L. M. S., & Nevin, J. A.
(1999). The effects of context and feedback on age
differences in spoken word recognition. The Journals of
Gerontology Series B: Psychological Sciences and Social
Sciences, 54(2), P125–P134.
Szekely, A., Jacobsen, T., D’Amico, S., Devescovi, A.,
Andonova, E., Herron, D., … Wicha, N. (2004). A new
on-line resource for psycholinguistic studies. Journal of
Memory and Language, 51(2), 247–250.
Wlotko, E. W., & Federmeier, K. D. (2012). Age‐related
changes in the impact of contextual strength on multiple
aspects of sentence comprehension. Psychophysiology,
49(6), 770-785.
Wlotko, E. W., Federmeier, K. D., & Kutas, M. (2012). To
predict or not to predict: Age-related differences in the use
of sentential context. Psychology and aging, 27(4), 975.
Zangl, R., & Fernald, A. (2007). Increasing flexibility in
children’s online processing of grammatical and nonce
determiners in fluent speech. Language Learning and
Development, 3(3), 199–231.
Zwitserlood, P., & Schriefers, H. (1995). Effects of sensory
information and processing time in spoken-word
recognition. Language and Cognitive Processes, 10(2),
121–136.
6
1198